Stock Trading using Fuzzy Support Vector Regression.
This project was done as part of the Neuro Evolution and Fuzzy Intelligence course in the NTU MSAI program.
Fuzzy models offer interpretability at the cost of accuracy. This project explores combining accurate Support Vector Regression (SVR) models using fuzzy membership functions in order to create an interpretable stock trading system. Additionally, model hyperparameters are tuned using Genetic Algorithm (GA). The method is compared with a vanilla SVR and a random walk.
See stock_trading_fsvr.ipynb for implementation details and results. See report.pdf for my presentation with more details about the project.